• We’re currently investigating an issue related to the forum theme and styling that is impacting page layout and visual formatting. The problem has been identified, and we are actively working on a resolution. There is no impact to user data or functionality, this is strictly a front-end display issue. We’ll post an update once the fix has been deployed. Thanks for your patience while we get this sorted.

[ComputerBase.de] Forza 7 Benchmark: Vega has more gasoline in the blood than Pascal

4K_shmoorK

Senior member
Graphics card benchmarks from Full HD to Ultra HD
The Radeon RX Vega 64, the Radeon RX Vega 64, turns the picture completely into Forza Motorsport 7 in the two lower resolutions tested: while the GeForce GTX 1080 Ti is usually drawn around the Radeon RX Vega 64 delivers 23 percent more images per second than the GeForce GTX 1080 Ti and the GeForce GTX 1080 - the real opponent for AMD's flagship - is hit 32 percent. Even the Radeon RX Vega 56 in Forza 7 is even faster than the T and 37 percent faster than the GeForce GTX 1070.

Even the duel Radeon RX 580 against the GeForce GTX 1060 decides for itself, however the distance with seven per cent is significantly lower. However, the GeForce GTX 970 must be clearly defeated: the Radeon R9 390 is 38 percent ahead. And the Radeon R9 380 is 36 percent faster than the GeForce GTX 960.
The higher the resolution, however, the more Nvidia is in motion. In 2.560 × 1.440, the Radeon RX Vega 64 still remains clear in the front. However, the lead to the GeForce GTX 1080 Ti shrinks to twelve percent and the GeForce GTX 1080 to 21 percent. In 3.840 × 2.160, the GeForce GTX 1080 Ti creates the first victory and delivers eight percent more FPS than the Radeon RX Vega 64. The AMD graphics card remains however still eleven percent faster than the GeForce GTX 1080. And the Radeon RX Vega 56 is 15 per cent faster than the GeForce GTX 1070. The Radeon R9 Fury X in the resolution, however, with four gigabytes of memory out, the title becomes unplayable. This only helps to reduce texture details or MSAA edge smoothing.

AMD also leads the Frametimes
AMD is also at the forefront of the Frametimes, although not as clearly as it is, because especially in 1.920 × 1.080, they are primarily limited by the CPU. All graphics cards cut in Full HD therefore very similar. Only the higher resolutions show that the frame distribution on the AMD graphics cards is somewhat better. For example, the Radeon RX Vega 64 delivers eight percent better frametimes in Ultra HD than the GeForce GTX 1080 Ti.

Nvidia confirms the backlog
The ranking in Forza 7 is very unusual. Nvidia has confirmed ComputerBase, however, that the results are so correct, so there is no problem with the system in the editorial regarding GeForce.

It is noticeable that the Nvidia graphics cards seem to be much earlier in a CPU limit than those of AMD: while for example the Radeon RX Vega 64 is synonymous in Full HD still good twelve percent faster than the Radeon RX Vega 56, is the Distance between the GeForce GTX 1080 and the GeForce GTX 1080 Ti only seven percent. And the GeForce GTX 1070 to Nvidia's flagship just 25 percent, in Ultra HD 39 percent - usually it is 60 percent.

The reason for AMD's very good performance does not seem to be a weakness of the competitor. Forza 7, the Vega architecture is obviously also very good. For example, the Radeon RX Vega 64 operates at a speed of 2,560 × 1,440 high 64 percent faster than the Radeon R9 Fury X, but on average the distance over other titles is just over 40 percent. With 73 percent to 56 percent lead, Vega is also above average with the Radeon RX 580. Both together lead to a great advantage for AMD. Closer to the GPU limit and without Vega is the advantage in the stroke exchange GTX 1060 against RX 580 significantly smaller.
https://www.computerbase.de/2017-09/forza-7-benchmark/

Very interesting. Wonder if this will be a growing trend amongst new DX12 titles?
 
May have something to do with the game's crappy CPU optimization. It hammers two cores, and barely uses the others. FH3 had the same problem until it was fixed in a patch nearly a year later, so it's surprising that FM7 kept the same CPU usage model that lead to stuttering and poor performance in FH3.

 
Last edited:
May have something to do with the game's crappy CPU optimization. It hammers two cores, and barely uses the others. FH3 had the same problem until it was fixed in a patch nearly a year later, so it's surprising that FM7 kept the same CPU usage model that lead to stuttering and poor performance in FH3.

yea that might explain the 4k performance, but still.. it's within 8% in GPU-limited scenario
 
Forza 7 seems to love Vega architecture. Surprisingly AMD GPUs perform worse at higher resolutions like 4K. Still there is room for optimizations i guess. Nvidia also has some optimization work left in this game. The gap between 1080 Ti and 1080 is too small and indicates there is room for performance improvements. I still think Vega launched with very poor launch drivers. I am looking forward to the yearly major driver release in December to see if there are driver based performance improvements for Vega across a wide range of games. RTG execution in 2017 has been abysmal. Hopefully 2018 execution from RTG is much better otherwise Nvidia will crush AMD GPU sales even more with Volta lineup.
 
yea that might explain the 4k performance, but still.. it's within 8% in GPU-limited scenario

You mean 1080p? I tried the demo, and I found the game to be very mild as far as taxing the hardware. For example, I can run at 4K 60FPS with 8x MSAA all settings maxed and easily maintain a locked 60 FPS. In fact, looking at the demo, my GPU usage was only in the 60s at these settings if I recall.

Most of that is due to the capped framerate in the demo, but still, the game's system requirements are tremendously overblown.
 
The gap between 1080 Ti and 1080 is too small and indicates there is room for performance improvements

This is why I think there is a CPU bottleneck with the core usage. Whoever's decision it was to use the same crappy thread model that caused so many issues in FH3 should be fired! Main loading one or two cores in a DX12 game is simply unacceptable, and reeks of incompetence.
 

Thanks for posting this. This game seems to have very poor CPU usage as well, as seen by the non existent scaling between the GTX 1080 and the 1080 Ti at 1080p In fact, the 1080 Ti doesn't really pull away significantly from the 1080 until 4K resolution.

It seems to reinforce the notion that NVidia's GPUs requires more CPU power than AMD's GPUs to hit maximum performance. And when a game has poor CPU usage, it seems to hurt NVidia GPUs more.

The YouTube video I posted above was done at 1440p as well, which should explain why the 1080 is ahead of the Vega 64.
 
Thanks for posting this. This game seems to have very poor CPU usage as well, as seen by the non existent scaling between the GTX 1080 and the 1080 Ti at 1080p In fact, the 1080 Ti doesn't really pull away significantly from the 1080 until 4K resolution.

It seems to reinforce the notion that NVidia's GPUs requires more CPU power than AMD's GPUs to hit maximum performance. And when a game has poor CPU usage, it seems to hurt NVidia GPUs more.

The YouTube video I posted above was done at 1440p as well, which should explain why the 1080 is ahead of the Vega 64.

Right. Nvidia's drivers have always done a lot of heavy lifting to contribute to the overall architecture speed. It remains that way today. It was clearly a good strategy for DX11 when the API was largely single threaded and could only scale to 2-3 cores efficiently. That left enough free cores for Nvidia's drivers to offload work to the CPU. Software is cheaper than hardware. I think this means that Nvidia at some point will be forced to move their driver function to hardware to compete. Still AMD crushing @ 1080p? It's not like DX12 didn't exist with Fiji. It's a complete reversal and that's why this is interesting to me. Vega definitely has teething issues, but it has shown maturity in these examples. My gut is that we are going to see more of this.
 
Don't even think its worthwhile testing early access games like pubg or any others, its usually developed by small teams that place performance optimizations last on the things to do list.

Only fully released games and even then after a patch or two would the performance metric be the accurate one.

But yeah, Vega is definitely much much better at DX12 and Vulkan, we see this even at BF1, Doom and other Dx12 games. Still though, most games are still DX11 and will continue to be so.

I see 2018 as a year where things might start to change in that department a lot more.
 
All AMD carda are punching above their weight here, Vega, 580, 390.. So this is unrelated to Vega per se, same situation with Dirt 4 actually.

But yeah, Vega is definitely much much better at DX12 and Vulkan, we see this even at BF1, Doom and other Dx12 games.
Nope, only Doom and BF1, the rest of DX12 either the 1080 wins or it's neck and neck with Vega 64 or slightly behind depending on the title.
I see 2018 as a year where things might start to change in that department a lot more.
Judging by 2017, less than 6 DX12 titles have been released, which is way less than 2016, developers are simply uninterested in DX12, and this disinterest increases every year.
 
All AMD carda are punching above their weight here, Vega, 580, 390.. So this is unrelated to Vega per se, same situation with Dirt 4 actually.


Nope, only Doom and BF1, the rest of DX12 either the 1080 wins or it's neck and neck with Vega 64 or slightly behind depending on the title.

Judging by 2017, less than 6 DX12 titles have been released, which is way less than 2016, developers are simply uninterested in DX12, and this disinterest increases every year.

Nvidia released their Forza 7 driver last week. The game is being reviewed right now with launch next week for the public. It doesn't seem likely Nvidia is going to have any big jumps in perf. You can't really punch above something you aren't capable of.

DX12 has waning interest because a dev could use Vulkan and support multiple OS.
 
May have something to do with the game's crappy CPU optimization. It hammers two cores, and barely uses the others. FH3 had the same problem until it was fixed in a patch nearly a year later, so it's surprising that FM7 kept the same CPU usage model that lead to stuttering and poor performance in FH3.


funny how this is a DX12 game by Microsoft and this exactly the opposite of what MS and DX12 promised to achieve.
 
How is dx12 interest waning? games take years to make and we had several dx12 games despite that. Its going to pick up as time goes on with lulls now and then.

Heavy CPU usage should be more of a problem for AMD gamers. I suspect a driver update or patch can fix this. maybe nvidia is doing their weirdness on the wrong cores.
 
Nvidia released their Forza 7 driver last week. The game is being reviewed right now with launch next week for the public. It doesn't seem likely Nvidia is going to have any big jumps in perf.
When the RX 580 is pushing better min fps than the 1080Ti, then you know there is some work yet to be done on NVIDIA's side.
How is dx12 interest waning? games take years to make and we had several dx12 games despite that. Its going to pick up as time goes on with lulls now and then.
The number of DX12 games dropped significantly in 2017 vs 2016.
 
There isnt anything nVidia can do about broken DX12 paths. It is up to the developer to fix it. Just look at Quantum Break.

And i dont see anything impressive about Vega either. It's only 70% faster than a GTX1060 in 4K...
 
BTW, a Turn10 developer stated that the reason for the 100% CPU usage on one or two cores is to reduce input latency, so it appears to be a design decision. But while multithreading might induce higher input latency, if done properly the penalty should be very minimal and be a nice trade off for a massive increase in framerates.
 
Forza 7 seems to love Vega architecture. Surprisingly AMD GPUs perform worse at higher resolutions like 4K. Still there is room for optimizations i guess. Nvidia also has some optimization work left in this game. The gap between 1080 Ti and 1080 is too small and indicates there is room for performance improvements. I still think Vega launched with very poor launch drivers. I am looking forward to the yearly major driver release in December to see if there are driver based performance improvements for Vega across a wide range of games. RTG execution in 2017 has been abysmal. Hopefully 2018 execution from RTG is much better otherwise Nvidia will crush AMD GPU sales even more with Volta lineup.

RTG's performance mostly hinges on them being able to counter their competitor's software ecosystem and I believe that D3D12 is their first step ...

Focusing on drivers won't be enough, RTG needs to expose more hardware features in a vendor agnostic gfx API and then get AAA game developers to use to use the said gfx API. Async compute, console style resource binding and the updates to the D3D12 spec such as DXIL shaders, custom resolve filters, barycentric coordinates is a good as well as stable foundation for RTG to start with. The next thing RTG should do when Microsoft is updating D3D12 again is to expose out of order rasterization (improves geometry performance!) and packed math in D3D12 ... (If RTG has to use proprietary technology to compete then they should since the current strategy doesn't work for them)

RTG has to defend their viability even if it means using what the community would consider underhanded tactics and RTG needs D3D12 to be their trojan horse ... (The point of competition is to always strive for a monopoly unless competitors decide to mutually cooperate but that's not realistic at all)
 
Where did you see this? 😵
here you go
index.php
 
Back
Top